Broadcom (AVGO.US) CEO: The Next Bottleneck for AI is Not Chips But Capital, Yet Many Investors Act Like Gamblers

According to Zhitong Finance APP, Broadcom (AVGO.US) President and CEO Hock Tan stated in Singapore on Monday that as the industry continues to ramp up infrastructure spending to support increasingly powerful models, capital—not chip supply—will become the biggest constraint on AI development. He also poured cold water on the influx of investors, saying that many of their strategies resemble gambling, with little understanding of their bets.

During the semiconductor roundtable at the inaugural DBS Asia Future Forward Conference, Hock Tan said, "Many investors have no idea what they're investing in." They "set extremely exaggerated expectations, only to realize in the end that they are almost just like being in a casino." Joining him on stage were Intel (INTC.US) CEO Lip-Bu Tan and DBS Group CEO Tan Su Shan, who served as moderator.

The Bottleneck Is Moving—And It's Now About Money

Hock Tan’s core judgment is: supply chain constraints will be resolved one by one, but capital will not. He stated at the event: "I think capital may be the most serious constraint." According to him, companies are gradually resolving wafer and memory shortages, but these bottlenecks do not disappear—they merely move elsewhere.

Reports indicate that he reflected on the past two years, having "seen a series of bottlenecks arise one after another, but we managed to overcome them," predicting that memory shortages "may not exist in a year," while wafer tightness is expected to ease by 2028 or 2029. After the sequence of "wafers, memory, power, land," capital becomes the biggest constraint.

Hock Tan pointed out that leading labs OpenAI and Anthropic have already demonstrated that their latest models, deployed in scenarios such as AI personal assistants, can generate substantial revenue, but financing will become the main bottleneck in the next phase.

He was even more blunt: the real driving force is not “selling shovels,” but whether leading AI companies can build valuable, profitable models to justify the current wave of infrastructure buildout. Value is moving up the industry chain; merely supplying hardware cannot support this boom.

"Still Losing Money": Endorsement and Reservations About OpenAI

On valuation issues, Hock Tan's stance contrasts with market rumors—he does not believe OpenAI is overvalued. According to reports, the company is seeking to raise at least $30 billion at a valuation of $1.4 trillion.

However, he interprets this number quite cautiously. Media quoted him as saying that OpenAI achieved about $50 billion in annualized revenue as of September, which suggests a feasible business model, but "there’s hardly any guarantee of long-term profitability." "They’re still losing money because they’ve spent too much on training," he said, "Frankly, at this stage, this doesn’t mean much—only that providing cutting-edge generative AI models presents a lot of potential for revenue growth."

He was even more unsparing about enterprise AI spending. Using a metaphor: model developers need to find customers who can actually make money through technology deployment, not rely on users who are currently "wasting money." "When you were a kid going to the arcade, you’d play, your parents would give you $10 in tokens, you’d spend it all in 10 minutes and learn nothing," he said. "Many companies today are just like that." He also questioned whether some companies are squandering money on AI and raised doubts about the long-term economic viability of loss-generating AI labs.

Broadcom: Those Saying Capital Is the Bottleneck Are Raising Money for Clients

Hock Tan is not a mere observer—Broadcom is deeply involved in the AI infrastructure financing chain, becoming one of its largest funding intermediaries.

According to reports from October 7, Broadcom is in early talks to arrange over $50 billion in debt financing for OpenAI, aimed at supporting OpenAI’s procurement of custom AI chips jointly developed by the two companies. Sources said Apollo Global Management and Blackstone have approached Broadcom, hoping to be lenders in the deal; negotiations are still in early stages and the final amount may vary. This funding could cover several gigawatts of computing power and is expected to close by the end of this year.

Broadcom is also a funder for Anthropic. According to Anthropic’s confidential IPO prospectus, Broadcom has agreed to provide up to $42 billion in financing to support Anthropic’s AI infrastructure expenditures; this amounts to about a third of Anthropic’s $125.2 billion, five-year TPU compute lease. The document also disclosed that Broadcom can appoint funding partners, and the related debt instruments can be converted into Anthropic shares. Anthropic expects to become Broadcom’s largest chip design customer by 2027—Broadcom itself expects AI semiconductor revenue of about $115 billion in FY2027 and around $230 billion in FY2028.

According to Proactive Investors, total Anthropic-related financing packages are reported to exceed $60 billion, including $42 billion in Broadcom-guaranteed senior debt and $18 billion in junior debt unbacked by Broadcom, led by Blackstone, with a syndication starting in October. Along with the $35 billion Broadcom-Apollo-Blackstone financing completed in June this year, and the proposed OpenAI deal, Broadcom’s involved funding arrangements have exceeded $145 billion. Additionally, Anthropic and xAI reportedly have committed to pay Broadcom about $161.2 billion for equipment leases by 2029.

In other words, the person saying “capital is the biggest bottleneck” is the one solving capital issues for clients with their own balance sheet. This is also the focus of market skepticism about AI deals.

Counter Evidence: An IPO Withdrawn, An Income Statement Controversy

Hock Tan’s warning has illustrative evidence. Firmus Grid, an Australian data center firm invested in by NVIDIA (NVDA.US) and supported by Blackstone and Coatue, withdrew its $5.5 billion IPO plan last week.

Firmus was set to issue at A$11 per share, implying about $30.4 billion in equity value—nearly three times its $10.5 billion valuation from less than three months ago. On October 9, the company announced it was withdrawing its Australian Securities Exchange listing application, citing recent market volatility, and would instead seek private fundraising, negotiating a round of $2–3 billion. The company’s projected FY2026 revenue is only $51 million, with just 46 MW completed out of 912 MW in its development pipeline—requiring investors to price a future that largely remains at the project planning and partnership agreement stage.

Just a week prior, there was another shock to the valuation anchor for AI deals. According to reports, OpenAI’s annualized revenue at the end of September was “close to $50 billion,” about $20 billion lower than the roughly $70 billion previously hinted to investors. The news triggered a collective tech stock dip on October 8: the Nasdaq Composite fell 1.3% to 27,193.34. However, the discrepancy was largely due to accounting differences—the higher figure was recalculated by OpenAI’s investors according to Anthropic's method, which counts customers' spending on its models via AWS, Google Cloud, etc., as revenue, while OpenAI does not include such spending.

In the same forum, Hock Tan’s comments subtly echoed this “accounting dispute”: he did not deny growth, but refused to equate growth with profitability.

High Interest Rates Are Locking Up the Chain

What’s making "capital as a bottleneck" more real is the price of funding. Currently, the 10-year US Treasury yield remains around 5.25%, not far from the 24-year high reached early last week.

Morgan Stanley estimates AI infrastructure will require about $1.5 trillion in external financing by 2028—making “access to funding” the primary variable for chip demand. Broadcom’s own financials do not yet show signs of pressure: for the third fiscal quarter of FY2026 (ending August 2), revenue was $29.6 billion, up 86% year-on-year, with AI semiconductor revenue of $16.7 billion (up 221%), accounting for 56% of total revenue. The company forecasts Q4 revenue at $34.8 billion and AI semiconductor revenue at $21.7 billion, raising FY2026 AI revenue guidance from $56 billion to $58 billion. As of the close on October 9, Broadcom was at $361.54 per share, with a market cap of about $1.73 trillion.

Another perspective at the event came from Intel. According to reports, Lip-Bu Tan believes AI’s impact outweighs that of the internet: the first wave is training, the next is inference and physical AI. He listed memory, supply chain, and power as constraints, predicting the future will not be “one chip fits all,” but a combination “XPU” architecture of GPUs, CPUs, and specialized accelerators. He also believes hardware will still be in short supply for the next two to three years, with memory chips likely to lose momentum first. Moderator Tan Su Shan opened the summit by stating the market is reassessing risk pricing for AI funding.

How Do Institutions See the Numbers?

The market’s evaluation of Broadcom’s role is clearly divided.

Seaport Research analyst Jay Goldberg said, “NVIDIA has committed a huge balance sheet, and Broadcom has had to follow.” He views Broadcom’s financing arrangements as a passive reaction to competitor strategies.

Rothschild & Co. Managing Partner Robert Leitao’s concerns are one step ahead: the market is “currently highly focused on one bet—whether these two companies can generate enough revenue to support all the financing.” Anthropic also acknowledged in its prospectus that Broadcom’s dual role as hardware supplier and financier poses potential conflicts of interest, possibly affecting the price and availability of its compute power.

Some institutions believe the market is overestimating the risks. According to a sell-side view cited by Pluang, Broadcom is rated “Buy” with a target price of $450, implying about 25% upside, arguing that the market overstates the risks of Broadcom’s $29 billion gap guarantee and $42 billion convertible financing. The company’s cash flow is robust, its exposure is limited, and its AI revenue guidance still points to $230 billion in FY2028 and earnings per share over $30.

Calculated at the October 9 closing price, Broadcom’s P/E ratio (TTM) is approaching 70 times. For a company that repeatedly calls “capital the biggest bottleneck” while simultaneously raising funds for clients, how much premium the market is willing to pay for that statement may be harder to answer than the statement itself.